Name bias in hiring is the well-documented tendency for a candidate’s name, as a signal of their ethnicity or gender, to change whether they get a callback, even when the rest of the CV is identical. It is one of the most rigorously measured effects in hiring research, because it has been tested directly: send out matched CVs that differ only in the name, and count who hears back. The results are consistent and uncomfortable. Names read as belonging to a majority group draw more callbacks than the same CV under a minority-group name, the gap has barely moved in decades, and it operates at exactly the moment a recruiter is skimming fast. Understanding the mechanism matters, because the fix follows directly from it: at the first pass, remove the name.
This is not about accusing recruiters of prejudice. Name bias is largely unconscious and it fires under time pressure, which is why well-intentioned people produce biased outcomes without noticing. The effect is a property of fast human judgement, not of bad character, and that is precisely why a structural fix works where good intentions do not.
The evidence, measured directly
Name bias is unusual in having causal evidence rather than correlation, because researchers have run controlled field experiments for decades.
The landmark hiring study sent otherwise identical CVs to real job openings, varying only the name at the top. The CVs with white-sounding names received roughly 50 percent more callbacks than the same CVs with African American-sounding names, a gap the authors noted was equivalent to several additional years of experience. That was one experiment; the pattern holds across many. A meta-analysis pooling 28 field experiments and more than 55,000 applications found white applicants received on average 36 percent more callbacks than African Americans and 24 percent more than Latinos, and, strikingly, that the level of discrimination had not declined in 25 years.
The effect is not confined to one country. When the UK government moved to name-blind recruitment, the Civil Service’s commitment cited evidence that applicants with white-sounding names were nearly twice as likely to get a callback, and adopted name-blind applications across the Civil Service and a group of major employers on exactly that basis. Different countries, different methods, same finding: the name changes the outcome.
What the numbers actually mean
It is worth being precise about what these studies show and do not show, because the effect is easy to both under- and over-state.
| The finding | What it means | What it does not mean |
|---|---|---|
| Identical CVs, different callbacks | The name alone changed the outcome | That every recruiter is consciously biased |
| The gap equals years of experience | It is large enough to change who gets hired | That it is the only factor in a decision |
| No decline over 25 years | Awareness alone has not fixed it | That nothing works; structural fixes do |
| Found across countries | It is a general effect, not local | That every workplace shows the same size |
The through-line is that name bias is real, large, and stubborn to good intentions, which is a specific and useful conclusion. It tells you that training people to be less biased has a poor track record against it, and that the reliable lever is to stop the signal reaching the decision in the first place.
Where in the process it does the most damage
Name bias operates hardest at the fast, high-volume first pass, and knowing that is what makes the fix targeted rather than vague.
When a recruiter skims a CV in about seven seconds, there is no time for deliberate, evidence-weighing judgement, so the fast, associative parts of cognition do the work, and those are exactly the parts a name activates. The slower stages, a proper read, an interview, a reference check, still carry bias, but the name is only one of many signals there and the pace allows reflection. The first pass is where a single signal, processed in a fraction of a second across hundreds of CVs, quietly reshapes who advances. That concentration is good news for the fix: you do not have to solve bias everywhere at once, you have to remove the name at the stage where it does the most, and cheapest, damage.
The fix that follows from the mechanism
Because the effect is a fast, unconscious response to a specific signal at a specific stage, the intervention is equally specific: hide the name, and the other identity signals that carry the same freight, during the first pass.
This is blind screening, and it works because it removes the input rather than trying to correct the response. Hiding the name, photo, age and often the specific school means the fast first pass runs on evidence, skills, experience, results, instead of on identity, and the broader practice of blind recruitment extends the same logic across the early stages. The mechanics of doing it on the CV pass specifically are in bias-aware blind CV screening. The key point is why it works where training fails: you cannot reliably train away an unconscious, split-second association, but you can stop the signal from ever reaching the reviewer.
| Approach to name bias | How it works | Track record |
|---|---|---|
| Unconscious-bias training | Asks reviewers to notice and correct it | Weak against a split-second effect |
| Blind the first pass | Removes the name before it is seen | Directly targets the mechanism |
| Structured criteria | Judges against fixed, job-relevant needs | Reduces room for the signal to matter |
| Diverse review panels | More perspectives on the decision | Helpful, but does not remove the signal |
What blinding the name does not fix
Removing the name is powerful precisely because it is targeted, but that also bounds what it can do, and overstating it is how the practice gets abandoned.
Blinding the first pass removes the name-based channel at the stage it matters most. It does not remove bias from interviews, where identity is visible again, and it does not fix criteria that themselves act as proxies, a specific university, an unbroken work history, which can reintroduce the same disparities under a different label. And it cannot change a biased applicant pool. Name-blinding is one precise intervention against one well-measured channel, most powerful combined with fair criteria and structured evaluation. That is not a weakness. An intervention that reliably removes a large, stubborn source of bias at the highest-impact stage is worth far more than a general programme that removes none of it.
A worked example
A mid-size company reviews its own hiring after noticing its shortlists rarely include candidates with non-Dutch names, despite a diverse applicant pool. Rather than assume malice or run another training session, it tests the pipeline the way the researchers do: it looks at the first-pass advance rates by name origin, and finds the gap opens at the CV screen, not at interview. The screen, done fast and by name-visible CVs, is where candidates are being lost.
The change is narrow and measurable. For the next few roles, the first pass is run name-blind, with names, photos and schools hidden while reviewers score the work and experience, and identity revealed only once a shortlist exists. The advance-rate gap at the first pass narrows sharply, because the signal that was driving it is no longer present when the fast decision is made. Nothing about the standard changed; the reviewers did not become better people over a fortnight. The company simply stopped feeding the name into the stage where it did the most damage.
What makes the example instructive is the method as much as the result: the company measured its own funnel, found where the gap opened, and applied the targeted fix at that stage. That is the realistic shape of countering name bias, a specific, checkable intervention at the first pass, not a vague commitment to fairness that never touches the mechanism.
Where a tool fits
Blinding names by hand across hundreds of CVs is the kind of task that quietly does not happen, which is where a tool earns its place. Zen Job CV offers bias-aware blind screening on the first pass: it hides names, photos, ages and schools while ranking candidates against your plain-language criteria, with the met-and-missed reasons attached, then you un-blind once a fair shortlist exists. Because the ranking is explainable rather than a black box, the blinding is not undone by hidden logic one layer down, and a person still reviews the shortlist and makes every decision.
The honest boundary is that a tool blinds the first pass; it cannot blind your interviews or fix criteria that act as proxies, and it works with whoever applied. It targets the one channel the evidence says matters most at the one stage where it does the most damage, which is exactly what the research says to do, and no more than that.
How to counter name bias in hiring
Accept what the evidence shows: name bias is real, large, unmoved by awareness alone, and concentrated at the fast first pass. Then act on the mechanism rather than on intentions by blinding the name, photo, age and specific school during the initial screen, so the first cut runs on evidence rather than identity. Pair it with fair, job-relevant criteria and structured evaluation to close the channels blinding does not reach, and treat interviews as a separate fairness problem where identity is visible again. Training people to try harder has a poor record against a split-second effect; removing the signal has a good one. Counter name bias where it is cheapest and most powerful to counter it, at the first pass, and build the rest of your fairness work around that.
Quick answers
What is name bias in hiring? It is the well-documented tendency for a candidate’s name, as a signal of ethnicity or gender, to change whether they get a callback even when the rest of the CV is identical. It has been measured directly with field experiments that send matched CVs differing only in the name, and it operates largely unconsciously and hardest at the fast first pass, which is why structural fixes work better than asking people to try harder.
How strong is the evidence for name bias? Strong and consistent. One landmark experiment found white-sounding names received about 50 percent more callbacks than identical CVs with African American-sounding names, and a meta-analysis of 28 studies and over 55,000 applications found white applicants got 36 percent more callbacks than African Americans, with no decline in 25 years. The UK Civil Service adopted name-blind hiring on similar evidence. The effect appears across countries and methods.
Does removing names from CVs actually reduce bias? For the name-based channel it targets, yes. Blinding removes the signal before it reaches the reviewer, which works where training struggles because you cannot reliably train away a split-second unconscious association. It is most effective at the fast first pass where name bias does the most damage. It does not remove all bias, since interviews and proxy criteria remain, but it directly counters the specific, measured channel.
Why doesn’t unconscious-bias training fix name bias? Because name bias is a fast, automatic association that fires in the seconds it takes to skim a CV, and asking people to notice and correct such an effect in real time has a weak track record. The reliable lever is structural: stop the signal reaching the decision by blinding the name at the first pass, rather than relying on reviewers to override an unconscious response under time pressure.
What are the limits of name-blind hiring? It removes one well-measured channel at the stage it matters most, but not all bias. It does not fix bias in interviews, where identity is visible again, or in criteria that act as proxies such as a specific university or an unbroken work history, and it cannot correct a biased applicant pool. Name-blinding is one precise, high-impact intervention, most powerful combined with fair criteria and structured evaluation, not a standalone cure.